The study was approved by the Institutional Review Board: Human Subjects Committee at the University of Minnesota Medical Center and Children's Hospitals and Clinics of Minnesota. Consent (and assent as appropriate) was obtained from children and their parent/guardian(s). We identified 723 living subjects, ages 9-18 years old, treated for cancer at the University of Minnesota Amplatz Children's Hospital and the Children's Hospitals & Clinics of Minnesota, in remission and surviving for ≥ 5 years after diagnosis of leukemia, central nervous system (CNS) tumors and solid tumors. Of these, 66 were not able to be contacted; of the remaining 657, 319 (49%) agreed to participate. 110 had leukemia, 127 solid tumors (i.e. sarcoma, renal, neuroblastoma, non-Hodgkin's lymphoma), and 82 CNS tumors (i.e. glial tumors, retinoblastoma, neuroectodermal tumors). 134 had a history of corticosteroid treatment, and 74 had a history of treatment with radiation (31 cranial radiation). Participants treated with hematopoietic cell transplantation (HCT) were excluded. There were no significant differences in age, sex, race, diagnosis, age at diagnosis and length of follow-up (time from diagnosis to study evaluation) between CCS participants and non-participants. A contemporary control group of 208 healthy siblings of CCS were recruited. Controls known to suffer from chronic illnesses including hypothyroidism and delayed puberty, or at risk for GH deficiency (i.e. height > 2 standard deviation (SD) below the mean and height velocity > 2 SD below the mean) were excluded from participation.
Following a 10-12 hour overnight fast, all participants underwent a physical examination (including Tanner staging [24
] by a trained study physician), height measured by wall mounted stadiometer (without shoes) to the nearest 0.1 cm and weight by electronic scale to the nearest 0.1 kg, and laboratory testing, including free thyroxine (free T4) by competitive immunoassay (CV 5.8-7.3%), thyroid stimulating hormone (TSH) (CV 4.9%), follicle stimulating hormone (FSH) (CV 5.8-6.1%), and insulin-like growth factor-1 (IGF-1) (CV 5.2-6.4%) by chemiluminescent immunoassay (Siemens Healthcare Diagnostics, Tarrytown, NY); and interleukin-6 (IL-6) (CV 14.5%) by ELISA (R&D Systems, Minneapolis, MN), adiponectin (CV 17.1%) and leptin (CV 13.7%) by 2-plex competitive immunoassay on the Luminex platform (Austin, TX) using bead sets from R&D Systems (Minneapolis, MN). GH stimulation test using clonidine and arginine was performed. GH deficiency was defined as a stimulated peak GH level less than 7 mcg/L, which is a more conservative cutoff than 10 mcg/L, which is frequently used in clinical practice to diagnose GH deficiency [25
]. Total body (not excluding head) BMD, posterior anterior lumbar spine (L2-L4), and body composition (percent body fat and lean body mass) were assessed by dual-energy x-ray absorptiometry (DXA) (G.E. Lunar Prodigy scanner; pediatric software version 9.3; Madison, WI, USA), and bone age by the Greulich and Pyle method [26
]. Dietary intake was evaluated using the Youth/Adolescent Questionnaire (YAQ) [27
]. Physical activity, including television/computer screen time, was assessed by the Modifiable Activity Questionnaire for Adolescents [29
]. The diagnoses of hypogonadism was made by either participant report of a history of diagnosed hypogonadism or for males an LH > 10 IU/L and testosterone below the lower end of reference range for pubertal stage and for females an FSH > 40 IU/L. The diagnosis of hypothyroidism was made by participant report, high TSH (> 5.0 uU/mL) and normal free T4, or low free T4 (< 0.8 ng/dL). Height and weight Z-scores were calculated based on 2000 Centers for Disease Control growth charts.
According to current International Society for Clinical Densitometry (ISCD) recommendation the terms osteoporosis should be limited to children with low BMD (Z-score ≤ -2) accompanied by fractures [30
]. Since very few patients met the definition of osteoporosis or low BMD in our study, yet a substantial proportion of patients had BMD below average, we defined a mild BMD deficit as a Z-score ≤ -1. This cutoff has been used in other studies describing bone deficits in CCS [22
]. The rationale for using this cutoff is that those with lower BMD Z-scores are likely to remain in the low end of the normal range [32
] and reach a lower peak BMD [11
] thus increasing their lifetime risk of osteoporosis and fracture. In addition, any decrease in BMD Z- score predisposed children to an increased risk of fracture [33
Descriptive statistics are expressed as frequencies and percents or mean ± standard error (SE), as appropriate. Regression models based on generalized estimating equations (GEE) with robust variance estimates were used to compare measures between CCS and the sibling control group with adjustments as noted in tables, to appropriately account for intra-family correlation. Among CCS, multivariable logistic regression was used to evaluate odds ratios (OR) for the associations between Z-score ≤ -1 of the whole body BMD and lumbar spine BMD with the following predictors: age at diagnosis, time since diagnosis, GH Status (GH deficient, not GH deficient), IGF-1 SDS (≤ -2, > - 2), percent body fat, lean body mass, body mass index (BMI), IL-6, adiponectin, leptin, calcium intake, vitamin D intake, zinc intake, omega-3 intake, protein intake, milk intake, fruit and vegetable intake, physical activity score, television/computer screen time, time elapsed since diagnosis, radiation and steroid exposure. All models were adjusted for sex, age-at-study, ethnicity (white-not-Hispanic, others), and pubertal Tanner stage.
The assumption of linearity for continuously valued factors was evaluated using Generalized Additive Models (GAM) [35
], and relevant categorical variables were created for those that were significantly non-linear at the alpha = 0.05 level, and, for adiponectin, leptin and IGF-1 to minimize undue influence by extremely large values [35
]. Age at study (whole body analysis only), IL-6, time elapsed since diagnosis (lumbar spine analysis only), and calcium were categorized per non-linear relationship with the outcomes. Category cut points were selected based on visual inspection of predicted curves from GAM models, and practical considerations regarding numbers of events per category. Because of correlation between percent body fat and leptin in both the whole body and lumbar spine models, separate multivariable models were developed including either leptin or percent body fat. Because of correlation between hypogonadism and hypothyroidism in the lumbar spine model, separate multivariable models were developed including either hypothyroidism or hypogonadism. Five models were evaluated: model 1: IGF-1 SDS, hypothyroidism and/or hypogonadism; model 2: lean body mass, percent fat mass (or leptin), television/computer screen time, physical activity score, years since diagnosis and IL-6; model 3: milk, protein, fruits/vegetable and daily total caloric intakes; model 4: protein, vitamin D, zinc, calcium, omega-3 and daily total caloric intakes; model 5: radiation and steroid exposure. All models were adjusted for age at study, sex, pubertal Tanner stage, and ethnicity. All p-values are two-sided and those < 0.05 were considered statistically significant, and those between 0.05 and 0.10 suggestive of association.